Free template
The one-page AI cost report, blank and yours
The structure our Snapshot produces, as a one-pager you can fill in by hand today. No account, no email, no attribution. If it is useful to you without us, take it.
1. What we spent
Every AI dollar in one place, by vendor and by category. If a line is on somebody's personal card, it still belongs here.
| Vendor | Category (seat / usage / infrastructure / other) | This month | Last month | Change |
|---|---|---|---|---|
2. What we can get back
Each recoverable line with a dollar figure and a proof level. A seat is unused when the licence has had no recorded activity for 60 days and is still being billed — write your definition down and do not move it.
| Finding | Dollars per month | Proof level (Counted / Compared / Estimated) | How it was measured | Owner |
|---|---|---|---|---|
3. What it might be earning
The estimated half, and it goes second on purpose. Show it as a range with the assumptions visible. If you cannot state the assumption, you do not have the number.
| Metric | Low estimate | High estimate | Assumptions used | What would change our mind |
|---|---|---|---|---|
4. What going faster cost
Beside every throughput claim, what going faster cost. Work redone soon after it shipped, bugs per change, reviews that were skipped, incidents. A speed number without its cost is not a result.
| What going faster cost | Before | After | Direction (better / worse) |
|---|---|---|---|
5. The rules this page was built under
Print these on the page, not in an appendix. They are what makes the rest of it survive a second reading.
- Each dollar is counted once, even when two findings touch the same seat
- Every figure states how it was measured
- Estimates render as ranges, never as a single number
- Definitions were fixed before the numbers were pulled
Proof Levels, for the column in section 2
- Counted
- read straight off a bill, a seat list or a usage report. No guessing. Two people pulling the same numbers would get the same answer.
- Compared
- the difference between two similar teams inside your own company, over the same period. Real signal, but a difference is not proof of a cause.
- Estimated
- worked out from assumptions you can see and change. Always shown as a range, never a single number.
The full method: how we grade a number.
Common questions
Is the AI cost report template really free?
Yes, and there is no email wall on it. The whole template renders on this page and the download is a plain markdown file. If it is useful to you without us, take it.
What goes in a one-page AI cost report?
Five sections: what you spent by vendor and category, what is recoverable with a proof level on each line, what the spend might be earning as a range, what going faster cost you in rework and defects, and the rules the page was built under. The order matters — counted first, estimated second.
Why does every line need a proof level?
Because a CFO's second question is always "how do you know". Proof Levels is three words: Counted, Compared, Estimated. A figure that cannot carry one of them is not ready to be in the report.
How much should we expect to be recoverable?
Across public benchmarks and our own sample data, 10 to 15% of seat and subscription spend is typically recoverable within 30 days. That denominator is deliberate: it is the recurring per-licence part of the bill, not the whole AI spend, because a token-heavy stack recovers less on seats.
Or have it filled in for you
Connect read-only sources and the same five sections come back populated from your own spend and usage, with a proof level on every line.